TauFlow: Dynamic Causal Constraint for Complexity-Adaptive Lightweight Segmentation

TL;DR

TauFlow is a lightweight medical image segmentation model that dynamically adapts to complex features, significantly improving accuracy and feature conflict mitigation on edge devices with extremely limited parameters.

Contribution

Introduces TauFlow with brain-inspired dynamic feature regulation and conflict mitigation modules for enhanced lightweight segmentation performance.

Findings

Reduces feature conflict rate from ~37% to ~9%.

Improves segmentation accuracy in extremely lightweight models.

Demonstrates effective boundary-background handling.

Abstract

Deploying lightweight medical image segmentation models on edge devices presents two major challenges: 1) efficiently handling the stark contrast between lesion boundaries and background regions, and 2) the sharp drop in accuracy that occurs when pursuing extremely lightweight designs (e.g., <0.5M parameters). To address these problems, this paper proposes TauFlow, a novel lightweight segmentation model. The core of TauFlow is a dynamic feature response strategy inspired by brain-like mechanisms. This is achieved through two key innovations: the Convolutional Long-Time Constant Cell (ConvLTC), which dynamically regulates the feature update rate to "slowly" process low-frequency backgrounds and "quickly" respond to high-frequency boundaries; and the STDP Self-Organizing Module, which significantly mitigates feature conflicts between the encoder and decoder, reducing the conflict rate from approximately 35%-40% to 8%-10%.